Skip to main content

Resultados relacionados

Resultados totales (Incluyendo duplicados): 7
Encontrada(s) 1 página(s)

Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

An analysis of the short utterance problem for speaker characterization

  • Viñals, Ignacio
  • Ortega, Alfonso
  • Miguel, Antonio
  • Lleida, Eduardo
Speaker characterization has always been conditioned by the length of the evaluated utterances. Despite performing well with large amounts of audio, significant degradations in performance are obtained when short utterances are considered. In this work we present an analysis of the short utterance problem providing an alternative point of view. From our perspective the performance in the evaluation of short utterances is highly influenced by the phonetic similarity between enrollment and test utterances. Both enrollment and test should contain similar phonemes to properly discriminate, being degraded otherwise. In this study we also interpret short utterances as incomplete long utterances where some acoustic units are either unbalanced or just missing. These missing units are responsible for the speaker representations to be unreliable. These unreliable representations are biased with respect to the reference counterparts, obtained from long utterances. These undesired shifts increase the intra-speaker variability, causing a significant loss of performance. According to our experiments, short utterances (3-60 s) can perform as accurate as if long utterances were involved by just reassuring the phonetic distributions. This analysis is determined by the current embedding extraction approach, based on the accumulation of local short-time information. Thus it is applicable to most of the state-of-the-art embeddings, including traditional i-vectors and Deep Neural Network (DNN) xvectors.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Supervector extraction for encoding speaker and phrase information with neural networks for text-dependent speaker verification

  • Mingote, Victoria
  • Miguel, Antonio
  • Ortega, Alfonso
  • Lleida, Eduardo
In this paper, we propose a new differentiable neural network with an alignment mechanism for text-dependent speaker verification. Unlike previous works, we do not extract the embedding of an utterance from the global average pooling of the temporal dimension. Our system replaces this reduction mechanism by a phonetic phrase alignment model to keep the temporal structure of each phrase since the phonetic information is relevant in the verification task. Moreover, we can apply a convolutional neural network as front-end, and, thanks to the alignment process being differentiable, we can train the network to produce a supervector for each utterance that will be discriminative to the speaker and the phrase simultaneously. This choice has the advantage that the supervector encodes the phrase and speaker information providing good performance in text-dependent speaker verification tasks. The verification process is performed using a basic similarity metric. The new model using alignment to produce supervectors was evaluated on the RSR2015-Part I database, providing competitive results compared to similar size networks that make use of the global average pooling to extract embeddings. Furthermore, we also evaluated this proposal on the RSR2015-Part II. To our knowledge, this system achieves the best published results obtained on this second part.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Unsupervised adaptation of PLDA models for broadcast diarization

  • Viñals, Ignacio
  • Ortega, Alfonso
  • Villalba, Jesús
  • Miguel, Antonio
  • Lleida, Eduardo
We present a novel model adaptation approach to deal with data variability for speaker diarization in a broadcast environment. Expensive human annotated data can be used to mitigate the domain mismatch by means of supervised model adaptation approaches. By contrast, we propose an unsupervised adaptation method which does not need for in-domain labeled data but only the recording that we are diarizing. We rely on an inner adaptation block which combines Agglomerative Hierarchical Clustering (AHC) and Mean-Shift (MS) clustering techniques with a Fully Bayesian Probabilistic Linear Discriminant Analysis (PLDA) to produce pseudo-speaker labels suitable for model adaptation. We propose multiple adaptation approaches based on this basic block, including unsupervised and semi-supervised. Our proposed solutions, analyzed with the Multi-Genre Broadcast 2015 (MGB) dataset, reported significant improvements (16% relative improvement) with respect to the baseline, also outperforming a supervised adaptation proposal with low resources (9% relative improvement). Furthermore, our proposed unsupervised adaptation is totally compatible with a supervised one. The joint use of both adaptation techniques (supervised and unsupervised) shows a 13% relative improvement with respect to only considering the supervised adaptation.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Multiclass audio segmentation based on recurrent neural networks for broadcast domain data

  • Gimeno, Pablo
  • Viñals, Ignacio
  • Ortega, Alfonso
  • Miguel, Antonio
  • Lleida, Eduardo
This paper presents a new approach based on recurrent neural networks (RNN) to the multiclass audio segmentation task whose goal is to classify an audio signal as speech, music, noise or a combination of these. The proposed system is based on the use of bidirectional long short-term Memory (BLSTM) networks to model temporal dependencies in the signal. The RNN is complemented by a resegmentation module, gaining long term stability by means of the tied state concept in hidden Markov models. We explore different neural architectures introducing temporal pooling layers to reduce the neural network output sampling rate. Our findings show that removing redundant temporal information is beneficial for the segmentation system showing a relative improvement close to 5%. Furthermore, this solution does not increase the number of parameters of the model and reduces the number of operations per second, allowing our system to achieve a real-time factor below 0.04 if running on CPU and below 0.03 if running on GPU. This new architecture combined with a data-agnostic data augmentation technique called mixup allows our system to achieve competitive results in both the Albayzín 2010 and 2012 evaluation datasets, presenting a relative improvement of 19.72% and 5.35% compared to the best results found in the literature for these databases.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Albayzin 2018 Evaluation: The IberSpeech-RTVE Challenge on Speech Technologies for Spanish Broadcast Media

  • Lleida, Eduardo
  • Ortega, Alfonso
  • Miguel, Antonio
  • Bazán-Gil, Virginia
  • Perez, Carmen
  • Gómez, Manuel
  • de Prada, Alberto
The IberSpeech-RTVE Challenge presented at IberSpeech 2018 is a new Albayzin evaluation series supported by the Spanish Thematic Network on Speech Technologies (Red Temática en Tecnologías del Habla (RTTH)). That series was focused on speech-to-text transcription, speaker diarization, and multimodal diarization of television programs. For this purpose, the Corporacion Radio Television Española (RTVE), the main public service broadcaster in Spain, and the RTVE Chair at the University of Zaragoza made more than 500 h of broadcast content and subtitles available for scientists. The dataset included about 20 programs of different kinds and topics produced and broadcast by RTVE between 2015 and 2018. The programs presented different challenges from the point of view of speech technologies such as: the diversity of Spanish accents, overlapping speech, spontaneous speech, acoustic variability, background noise, or specific vocabulary. This paper describes the database and the evaluation process and summarizes the results obtained.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Optimization of the area under the ROC curve using neural network supervectors for text-dependent speaker verification

  • Mingote, V.
  • Miguel, A.
  • Ortega, A.
  • Lleida, E.
This paper explores two techniques to improve the performance of text-dependent speaker verification systems based on deep neural networks. Firstly, we propose a general alignment mechanism to keep the temporal structure of each phrase and obtain a supervector with the speaker and phrase information, since both are relevant for a text-dependent verification. As we show, it is possible to use different alignment techniques to replace the global average pooling providing significant gains in performance. Moreover, we also present a novel Back-end approach to train a neural network for detection tasks by optimizing the Area Under the Curve (AUC) as an alternative to the usual triplet loss function, so the system is end-to-end, with a cost function close to our desired measure of performance. As we can see in the experimental section, this approach improves the system performance, since our triplet neural network based on an approximation of the AUC (aAUC) learns how to discriminate between pairs of examples from the same identity and pairs of different identities. The different alignment techniques to produce supervectors in addition to the new Back-end approach were tested on the RSR2015-Part I and RSR2015-Part II database for text-dependent speaker verification, providing competitive results compared to similar size networks using the global average pooling to extract supervectors and using a simple Back-end or triplet loss training.
Zaguán. Repositorio Digital de la Universidad de Zaragoza
Publication Artículo científico (JournalArticle)

Knowledge Distillation and Random Erasing Data Augmentation for Text-Dependent Speaker Verification

  • Mingote, Victoria
  • Miguel, Antonio
  • Ribas, Dayana
  • Ortega, Alfonso
  • Lleida, Eduardo
This paper explores the Knowledge Distillation (KD) approach and a data augmentation technique to improve the generalization ability and robustness of text-dependent speaker verification (SV) systems. The KD method consists of two neural networks, known as Teacher and Student, where the student is trained to replicate the predictions from the teacher, so it learns their variability during the training process. To provide robustness to the distillation process, we apply Random Erasing (RE), a data augmentation technique which was created to improve the generalization ability of the neural networks. We have developed two alternatives of the combination of KD and RE, which, produce a more robust system with better performance, since the student network can learn from teacher predictions of data not existing in the original dataset. All alternatives were tested on RSR2015-Part I database, where the proposed variants outperform reference system based on a single network using RE.